How to go about learning all of this? 1st: Start with the serving engine view – vLLM: PagedAttention, continuous batching, prefix caching, CUDA graphs – SGLang: RadixAttention/prefix reuse, speculative decoding, MoE, structured/agent workloads – TensorRT-LLM: NVIDIA peak
RESEARCH
-
AI Referees in Beach Volleyball: Enhancing Game or Changing Its Soul?
By
–
#AI Referees the Sand: Real-Time Beach Volleyball Analysis — Enhancing the Game or Changing Its Soul?
— Ronald van Loon (@Ronald_vanLoon) 8 juin 2026
by @measure_plan#ArtificialIntelligence #MachineLearning #ML pic.twitter.com/rlLWq2ZKgc#AI Referees the Sand: Real-Time Beach Volleyball Analysis — Enhancing the Game or Changing Its Soul?
by @measure_plan #ArtificialIntelligence #MachineLearning #ML -
Gary Marcus: Not close to RSI despite Anthropic’s hints
By
–
tl;dr: we aren’t close to RSI, regardless of the hints IPO-bound Anthropic tried to drop last week.
-
Request for long context benchmarks for decode and prefill
By
–
Give me long context benches for decode and prefill
-
How AI uses the brain’s oldest cognitive bias
By
–
I have published an episode on @ivoox: "#1141: How AI uses the oldest cognitive bias of the human brain #podcast"
-
Lab will take care of small specialized models as future
By
–
My lab will take care of this I had a stream in October 2024 saying small and specialized models are the future that I need to find
-
From C to Theano: The Evolution of Neural Network Frameworks
By
–
I wrote my first neural networks in pure C, then in Matlab, then in NumPy, before eventually moving to Theano. Since then, I have seen and tried practically every neural network framework ever developed. Some are bad, some are good. The good
-
AI that creates AI: Launch of the RSI Lab
By
–
AI that creates AI that creates AI: Launch of the RSI Lab https://
sakana.ai/rsi-lab-jp/ Sakana AI launches in Tokyo a research group dedicated to Recursive Self-Improvement (RSI), named 'RSI Lab'. RSI is the mechanism by which AI creates AI. -
Samsung MeKi: Memory-Based Expert Knowledge Injection for LLMs
By
–
Want to scale LLMs without skyrocketing compute costs? Samsung presents MeKi: Memory-based Expert Knowledge Injection. Instead of making models bigger to learn everything, MeKi gives LLMs a dynamic memory bank of expert knowledge. Think of it as a cheat sheet the model can
-
Why Geoffrey Hinton called it Dark Knowledge in CIFAR meetings
By
–
I’m wondering why @geoffreyhinton called it Dark Knowledge in the earlier @CIFAR_News meetings??